MoGCN: A Multi-Omics Integration Method Based on Graph Convolutional Network for Cancer Subtype Analysis.

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Large-scale omics datasets are rapidly accumulating.
• Integrating multi-omics data for cancer research presents significant challenges.
• Machine learning methods are increasingly used for cancer molecular and clinical feature characterization.

Purpose of the Study:

• To develop a multi-omics integration model, MoGCN, for cancer subtype classification and analysis.
• To address the challenges of integrating heterogeneous omics data using graph convolutional networks (GCN).
• To enhance biological knowledge discovery and clinical diagnosis through interpretable classification.

Main Methods:

• Developed MoGCN, a multi-omics integration model based on GCN.
• Utilized autoencoder (AE) for dimensionality reduction and similarity network fusion (SNF) for patient similarity network (PSN) construction.
• Integrated genomic, transcriptomic, and proteomic data from TCGA breast invasive carcinoma (BRCA) and pan-kidney cancer datasets.

Main Results:

• MoGCN achieved superior accuracy in cancer subtype classification compared to existing algorithms.
• The model successfully extracted significant features from each omics layer and identified candidate functional molecules.
• Network visualization provided clinically intuitive diagnostic insights.

Conclusions:

• MoGCN demonstrates robust performance in heterogeneous data integration and provides interpretable classification results.
• The model shows significant potential for biomarker identification and clinical diagnosis in cancer research.
• MoGCN offers a powerful tool for analyzing complex multi-omics data for improved cancer understanding.